Cleveland State University EngagedScholarship@CSU
Sociology & Criminology Faculty Publications Sociology & Criminology Department
7-2009
Bond to Society, Collectivism, and Conformity: A Comparative Study of Japanese and American College Students.
Miyuki Fukushima Cleveland State University, [email protected]
Susan F. Sharp University of Oklahoma Norman Campus
Emiko Kobayashi Kanazawa University
Follow this and additional works at: https://engagedscholarship.csuohio.edu/clsoc_crim_facpub
Part of the Race and Ethnicity Commons, Social Control, Law, Crime, and Deviance Commons, and the Sociology of Culture Commons How does access to this work benefit ou?y Let us know! Publisher's Statement This is an Author’s Accepted Manuscript of an article published in Deviant Behavior, July 2009, available online: http://www.tandfonline.com/10.1080/01639620802296212.
Repository Citation Fukushima, Miyuki; Sharp, Susan F.; and Kobayashi, Emiko, "Bond to Society, Collectivism, and Conformity: A Comparative Study of Japanese and American College Students." (2009). Sociology & Criminology Faculty Publications. 101. https://engagedscholarship.csuohio.edu/clsoc_crim_facpub/101
This Article is brought to you for free and open access by the Sociology & Criminology Department at EngagedScholarship@CSU. It has been accepted for inclusion in Sociology & Criminology Faculty Publications by an authorized administrator of EngagedScholarship@CSU. For more information, please contact [email protected]. bond to society, collectivism, and conformity: a comparative study of Japanese and American college students
Miyuki Fukushima and Susan F. Sharp Department of Sociology, University of Oklahoma, Normal, Oklahoma, USA Emiko Kobayashi Foreign Language Institute, Kanazawa University, Kanazawa, Japan
An argument is developed that the purported collectivism in Japanese society generates stronger social bonds in Japan than in the more individualistic United States, which might then explain the lower level of deviance often found in Japan. We test this using survey data from samples of Japanese and American college students on measures of deviance and social bonds. Results indicate that Japanese students engage in significantly less deviance than Americans, and
Funding for the surveys reported was provided by the College of Arts and Sciences, Uni- versity of Oklahoma and by Grant-in-Aid for Scientific Research #16730274 from the Japa- nese Ministry of Education, Culture, Sports, Science and Technology (Emiko Kobayashi, investigator). An earlier version of this article was presented at the Annual Meeting of the Southwes- tern Social Science Association in 2005. We are most grateful to Dr. Harold G. Grasmick for his contributions to collecting the U.S. data and his invaluable input into the manuscript. We also thank anonymous reviewers of this journal for their insightful comments and critiques. Address correspondence to Miyuki Fukushima, Department of Sociology, University of Oklahoma, 780 Van Vleet Oval, Kaufman Hall, Room 342, Norman, OK 73019, USA. E-mail: [email protected] although variables from Hirschi’s (1969) social control theory behave similarly across cultures as predictors of deviance, the theory failed to account for the lower level of deviance among Japanese.
INTRODUCTION The contrast in crime rates between Japan and the United States carries salience for comparative sociology, as Japan is a non-Western country that is relatively comparable to the United States in economic development. Scholars have repeat- edly noted evidence of much lower crime rates in Japan than in the United States and other Western nations (see Adler 1983; Braithwaite 1989; Dussich et al. 2001; Gruszczynska 2002; Hendry 1989; Komiya 1999; Roberts and Lafree 2001, 2004). However, only a handful of studies (e.g., Grasmick and Kobayashi 2002; Kobayashi and Grasmick 2002; Kobaya- shi et al. 2001; Vazsonyi et al. 2004) have used comparable self-report data demonstrating lower levels of deviance among Japanese than Americans or used conventional theories of deviance to attempt to explain this difference. The present study employs identical self-reported deviance measures administered to samples of Japanese and American college students in April 2003. Although our study deals with the relatively nondeviant group of college students, we none- theless found the expected difference in the levels of deviance between the Japanese and American samples—the level of deviance measured in this study is lower among the Japanese students. Our research is guided by the literature on collectivis- tic and individualistic culture in comparative sociology and psychology, which we argue can be tied to Travis Hirschi’s (1969) social control theory of deviance as an explanation of compliance with norms. We hypothesize that social bonds will be stronger in a collectivistic culture like Japan, leading to higher levels of compliance with norms. Our survey contains direct measures of variables from Hirschi’s social control theory to test the hypothesis that Japanese engage in less deviant behavior because they are more strongly bonded than Americans to conventional society. Hirschi (1969) argues that his social control theory is applicable in explaining deviant behaviors across groups and cultures and should explain differences in levels of deviance between groups (see also Chapple et al. 2005; for a similar argument concerning Gottfredson and Hirschi’s more recent self-control theory, see Tittle and Botchkovar 2005)—the two important issues for comparative criminol- ogy originally raised by feminist scholars. Namely, can crim- inological theory (1) explain equally well delinquency= deviance of different groups and (2) account for the differ- ences in the level of delinquency=deviance of groups? The first issue pertains to the generalizability or applicability of a theory of deviance across diverse groups, whereas the sec- ond issue pertains to how well a theory accounts for the gap in deviance between groups. Hirschi (1969) argues that as a universal theory of deviance, his social control theory should account for both of these issues. Some empirical studies testing Hirschi’s social control theory of deviance, or part of the theory, beyond Western culture have been reported. Without exception, however, all these studies test the theory within only non-Western cultures and thus lack a comparison group from a Western culture. These studies can demonstrate only whether social bond variables affect deviance beyond Western culture or whether Hirschi’s social control theory is applicable in non-Western culture (i.e., the first issue). They cannot, how- ever, demonstrate whether social bond variables can account for differences in levels of deviance across cultures (i.e., the second issue). Using comparable data from Japa- nese and American college students on deviance and social bonds, our main objective is to examine the second issue, that is, can Hirschi’s social control theory account for the expected gap in deviance between Japanese and American students?
HIRSCHI’S SOCIAL CONTROL THEORY Travis Hirschi (1969) introduced a refined version of control theory into the field of deviance, building on existing control theories (cf., Reckless 1967), and focusing on the social bond. Hirschi draws a sharp distinction between control the- ory, which emphasizes constraints against deviance, and other theories that concentrate on factors that push people into deviance, such as Merton’s anomie theory (see also Gottfredson and Hirschi 1990; Kornhauser 1978). For control theory, all actors are assumed to be equally motivated to transgress, but people vary in the constraints they experience that more or less prevent them from acting on this universal motivation to engage in acts of force and fraud in pursuit of their self-interest; thus, control theory is about these constraints. According to Hirschi (1969), the controls or constraints acting on individuals and preventing deviance are found in their bonds to conventional society. Deviance occurs when ‘‘[an] individual’s bond to society is weak or broken’’ (Hirschi 1969:16). In other words, the stronger an individual’s bond to society, the less likely the individual is to deviate from societal norms. Hirschi (1969) identifies four elements of the social bond: attachment, commitment, involvement, and belief. The ele- ment of attachment is the ‘‘bond of affection’’ to conventional persons and institutions. Commitment refers to the ‘‘stakes in conformity that are built up by pursuit of, and by a desire to achieve, conventional goals’’ (Hirschi 1969:162). Involve- ment, as argued by Hirschi (1969:185) is ‘‘the most obviously related to delinquent behavior,’’ for the more time people spend in conventional activities, the less time they have for delinquent and criminal activities. Finally, for Hirschi (1969:198), the element of belief in the legitimacy of law is related to deviance in that the ‘‘absence of (effective) beliefs’’ that forbid deviance, free one to engage in it. A handful of studies test Hirschi’s version of social control theory in Asian societies (Hwang and Akers 2003; Miller 1992; Saito 2002; Shoemaker 1994; Tanioka and Glaser 1991; Wang et al. 2002; Zhang and Messner 1995, 1996). Generally, but with exceptions, they find some evidence that in Asian samples at least some of the elements of the social bond are linked to lower levels of self-reported deviance. Notable anomalies are Zhang and Messner (1996) in China and Hwang and Akers (2003) in South Korea, who find that attachment to parents is not important to explaining self- reported deviance in these societies. In addition, in the Hwang and Akers (2003) study, the effect of attachment to peers on deviance is found to be positive, contrary to Hirschi’s (1969) prediction. The objective with our unique dataset is different from previous studies testing the theory among Asian samples. In comparable samples of Japanese and American college students, we measured all of the elements of the social bond, along with self-reported deviance. We expect to find lower levels of deviance among the Japanese, and we expect this to be the result of stronger social bonds among them. Though no extant studies empirically examined a direct link between individualism–collectivism and the social bond, in the next section, we point to the literature on collectivism as the rationale for our central hypothesis that Japanese youth engage in less deviant behavior than Americans because they are more strongly bonded to conventional society.
COLLECTIVISM–INDIVISUALISM The idea that Japan, in contrast to the United States, is more collectivistic emerged in writings such as Lowell’s (1888) The Soul of the Far East, published shortly after the reopening of Japan in 1854. The most widely known statement on this topic no doubt is Benedict’s (1946) The Chrysanthemum and the Sword (see also Abegglen 1958). The idea that Japan tends more toward collectivism, whereas the United States tends toward individualism also gained popularity even among Japanese scholars (see Doi 1971, 1985; Kawasaki 1969; Nakane 1970). In the 1980s, researchers, especially Hofstede (1980; Hofstede and Bond 1984; see also Hofstede 1991) in the field of cross-cultural psychology began developing measures of individualism–collectivism and using these measures to collect data from diverse cultures (see also Triandis 1988, 1990, 1995). In his seminal empirical study of IBM workers in different countries, Hofstede (1980; see also Hofstede and Hofstede 2004) reported that American workers in their work setting were significantly more indivi- dualistic on his measure than were Japanese workers. Since Hofstede’s original formulation and research, others have developed modifications of his basic idea. In the tradition of theory and research on ‘‘value orientations’’ (e.g., Ball- Rokeach et al. 1984; Rokeach 1973), Schwartz (1990, 1992) conceptualized individualism–collectivism as more global ‘‘value orientations.’’ Markus and Kitayama (1991; see also DeGooyer 1992) linked the idea to self-concept literature, distinguishing between an ‘‘independent self-concept’’ (i.e., individualism) versus an ‘‘interdependent self-concept’’ (i.e., collectivism). Throughout this literature, Japan, compared to the United States, is found to be collectivistic, and Japanese, compared to Americans, are considered to have more inter- dependent self-concepts. Furthermore, using questionnaire items from refinements of Schwartz’s (1990, 1992) and Markus and Kitayama’s (1991) scales, Gudykunst et al. (1996) also reported that Japanese college students scored higher than Americans on measures of collectivistic values (e.g., harmony with others, observing rites and social rituals) and an interdependent self- concept (e.g., ‘‘I consult with others before making important decisions,’’ ‘‘I will sacrifice my self-interest for the benefit of my group’’). Our contention is that collectivism implies a stronger bond to society, and, if Hirschi’s (1969) theory is correct, this should lead to less deviance in collectivistic societies like Japan than in more individualistic societies like the United States. The similarity between Hirschi’s refined social control theory that focuses on the individual’s bond to society (social bond) and the individualistic–collectivism dimension derives from an identical emphasis placed on the relationship between the individual and the collective society. Individualism means placing a value on individual identity over group identity, individual needs and rights over group obligations, and individual pleasure over adherence to group norms (for a discussion of individualism in the United States, see Bellah et al. 1985). Collectivism, on the other hand, grants priority to group identity over individual identity, shared in-group beliefs over unique individual beliefs, and cooperation with in-group members over maximizing indivi- dual outcomes (see Gudykunst et al. 1996). Triandis (1988) states that members of individualist cultures are character- ized by hedonism and self-seeking attitudes, suggesting weak bonds to society. Hence, we argue that collectivist cul- tures should be more effective in fostering social controls or constraints against deviant behavior while also fostering conformity among members of society. Indeed, cross-cultural scholars posit that members of col- lectivist cultures, such as Japan’s, tend to place a higher value on family, family relationships, and on relationships with others in the society, while members of individualistic cultures, such as America’s, tend to place a higher value on independence and autonomy (see Gudykunst et al. 1996; Hofstede 1980, 1984, 1991, 2001; Matsumoto and Juang 2004; Matsumoto et al. 1996, 1997; Triandis 1988, 1999). In other words, using Hirschi’s (1969) terminology, we might argue that members of collectivist cultures are more strongly attached to parents, peers, teachers, and schools. Furthermore, the literature on collectivism suggests that the ‘‘stake in conformity,’’ or the element of commitment, might also be higher in collectivist cultures than in individu- alist cultures. The greater social pressure to conform and, thus, greater potential costs (stake) from deviant behavior in collectivist cultures, should promote a higher level of commitment among their members, compared to individual- ist cultures, particularly in the United States where confor- mity is, to some extent, considered a negative attribute (Matsumoto and Juang 2004). In addition, the stronger attachment to others, coupled with the greater commitment to conventional goals, should then produce a higher level of involvement in conventional activities in collectivist cultures. Finally, members of collectivistic cultures tend to value ‘‘norms, duties, and obligations’’ relative to individual factors, while the members of individualistic cultures tend to value personal emotions, attitudes, and needs relative to social factors (Triandis 1999:128). It would follow, then, from Hirschi’s (1969) formulation that collectivist cultures promote a stronger bond to society by generating a stronger belief in the legitimacy of the law. The idea that the concept of the Hirschi’s (1969) social bond might overlap with the concept of collectivistic versus individualist culture in cross-cultural psychology does not derive directly from his formulation of social control theory. To be clear, the concept of collectivistic–individualistic cul- ture explains cultural variations across different groups. On the other hand, although Hirschi intended his theory to be a generalizable, universal theory to be applied to all and any groups, Hirschi’s social control theory is not an explana- tion of between-group cultural variations (e.g., individual- ism–collectivism) in deviance. Furthermore, there is no extant study that directly tests the link between individual- ism–collectivism and social bonds, from which we could base our assumption that collectivistic culture has higher social bonds compared to individualistic cultures. Neverthe- less, theories and research in this area outlined here lead us to reasonably infer that social bonds will be stronger in col- lectivistic than in individualistic cultures, which then would explain the lower level of self-reported deviance we expected and found among Japanese college students com- pared to Americans. Thus, our research tests the hypothesis that Japanese college students score lower than Americans on self-reported deviance because Japanese score higher on the elements of the social bond outlined by Hirschi (1969)—attachment, commitment, involvement, and belief in the legitimacy of the law.
METHODS Sample The data to test the hypotheses come from a survey conducted by two of the authors. The American sample consists of 505 students from a Southwestern state university who were enrolled in an introductory sociology class. The Japanese sample consists of 442 students who were enrolled in sophomore- level classes at a comparable national university in Japan, to which one of the Japanese authors had access. The ques- tionnaire initially was written in English and then was carefully translated and pre-tested in Japan by one of the Japanese authors. Data were collected in the two universities at the same time in April 2003. Because of the differences in academic calendars, this means that most of the American students were just finishing their freshman year, whereas the Japanese students had just began their sophomore year. Gathering comparable survey data from college students in the two countries is not a simple task. Several issues had to be addressed concerning differences in social structure and culture, differences that affected both sample selection and measurement. To achieve comparability in ways that are important for the analysis, some cases, as explained later, had to be omitted. The final analysis to be reported is based on 433 of the Japanese students and 369 of theAmericanstudents. Data were obtained from students in two public universi- ties—one in the United States and one in Japan. In the United States, this was a major state university in the South- west to which the American authors had access. The total enrollment (graduate and undergraduate) is approximately 22,000, and the university is within the boundaries of a metropolitan area of about 1.1 million inhabitants that also contains the state’s capital. Selecting a comparable Japanese university required sev- eral decisions. First, the term ‘‘state university’’ can be mis- leading in Japan. Each of the 47 Japanese prefectures (equivalent to states in the United States) has several prefec- ture- and city-financed universities, but these generally are not comparable to the major flagship universities in states in the United States. Rather, they more closely resemble regional universities within states in the United States. In addition to these Japanese prefecture- and city-financed uni- versities, however, each prefecture has at least one national university funded by the Japanese Ministry of Education, Culture, Sports, Science and Technology. As of 2003, the year our data were gathered, there were 100 of these national uni- versities, which are the major and prestigious, and most selective public universities in the country. These Japanese national universities are the ones most comparable to the major state universities in the United States. Consequently, the Japanese data were collected at a national university. To select a particular national university in Japan, similar- ity to the U.S. metropolitan area in the Southwest and to the size of the university in the United States, were important considerations. We chose a national university in a prefec- ture that did not contain the largest Japanese metropolitan areas like Tokyo, Yokohama, and Osaka. Nevertheless, it is located within a metropolitan area of substantial size (2.2 million inhabitants) that contains the prefecture’s capital city. The university selected has an enrollment of approxi- mately 16,500 undergraduate and graduate students. Thus, both the American and the Japanese universities are in large (but not the largest) metropolitan areas that include the state= prefecture capital. Although the American state university has a somewhat larger enrollment than the Japanese national university, neither is among the largest such university in their country. We also had to address the wide discrepancy between Japanese national and American state universities in racial and ethnic diversity, a discrepancy so wide that ‘‘minority group’’ status could not be a variable in our analysis. There are essentially no ‘‘minority group’’ members in Japanese national universities. Had we tried to include a variable for minority group status, all Japanese respondents would have been in the same category. Consequently, our plan was to use only the questionnaires completed by Caucasian stu- dents in the United States, excluding those who were self- identified minority group members. Because the American university’s enrollment is 74% Caucasian, a total sample of 600 would have yielded about 450 Caucasians. However, one of the sections of Introduction to Sociology from which we intended to gather data became problematic, and our total U.S. sample was 505, of whom 369 (i.e., 73%, com- pared to 74% in the university according to official figures) identified themselves as ‘‘white.’’ All others have been excluded from the analysis we report. The absence of ‘‘minorities’’ in Japanese national universi- ties merits explanation. The Ministry of Public Management, Home Affairs, Posts and Telecommunications, Japan (1995) reports populations of people of non-Japanese origin, primar- ily Koreans, Chinese, Brazilians, and Filipinos (see also Sugimoto 2003). However, a series of laws over time has made it difficult, often impossible, for such people to obtain Japanese citizenship (Lie 2001:83–110). Even through marriage to a Japanese citizen, citizenship is not guaranteed. Furthermore, those residents without Japanese citizenship, with rare exceptions, are not eligible to attend Japanese national universities. The Japanese government does not report figures on the number of non-Japanese permanent residents enrolled in national universities, but the number surely is low. A conse- quence of this policy of exclusion is that 98% of the students in our Japanese sample do not identify themselves as having non-Japanese origin. The remaining 2% (nine students) are Japanese residents who are not Japanese citizens or non- resident students from other Asian countries. These nine are excluded from the analysis, leaving 433 useable cases in Japan. In the analysis, these will be compared to the 369 whites in the U.S. sample. In both countries, all students who were in class the day the data were collected agreed to complete the questionnaire. Measures Japan Japan is a dummy variable coded 1 for Japanese and 0 for American students. The variable has a mean of .540 and standard deviation of .498. We expect the variable Japan to have an inverse effect on our measure of self-reported deviance described later, even when other variables, also described later, are controlled. Self-Reported Deviance The dependent variable consists of 13 items measuring various non–drug related deviant behaviors ranging from ‘‘Hurt someone badly enough that they needed bandages or a doctor’’ to ‘‘Cheated in school to get a better grade.’’ Students were asked to indicate how often they have engaged in the 13 deviant behaviors in the past year using a five-point Likert scale with responses ranging from ‘‘Never’’ to ‘‘Almost always.’’ Items are coded so that a higher value indicates a higher level of deviance, with values ranging from 0 to 4. Before creating a single deviance scale from the combined samples, we assessed the possibility that the dimensionality of deviance is not the same in the two countries. In both cases, however, the eigenvalues indicated unidimensional- ity, and the reliability coefficients were both identically high (.763 in Japan and .812 in the United States). These analyses suggest that we can be justified in creating a single deviance scale from the combined samples. A prin- cipal components analysis for the 13 items indicates a single factor. Cronbach’s alpha for the linear composite of z-scores of the 13 items is .79. Reliability could not be improved by eliminating any of the items. The deviance scale was created by summing the z-score transformations of the all items, producing a deviance scale for the combined sample with a mean of 0 and standard deviation of 7.07. Control Variables Four control variables generally thought to be relevant in studies of deviance are included in the analyses, including respondent’s age, gender, parental socioeconomic status, and family structure. A preliminary analysis of the age distribution of the combined sample shows that the distribu- tion is positively skewed (sk ¼þ4.26). This is due to a small number of outliers in the American sample who are much older than average college students (e.g., sk for the American sample is þ3.52 and for the Japanese sample is þ1.81). In order to reduce the skewness of the distribution of age, par- ticularly among the American sample, the log of the variable age for the combined sample is created by taking the natural log of each student’s actual age. The new distribution of log- age shows a slight reduction in skewness (sk ¼þ2.93) with a mean of 1.29 (SD ¼ .023). Respondents’ gender is a dummy variable called Male with males coded as 1 and females coded as 0. The variable Male has a mean of .58 (SD ¼ .49). It should be noted that the two samples differ significantly in percent male (43% in the United States and 71% in Japan). The figures from the Japanese Ministry of Education, Culture, Sports, Science, and Technology (2003) indicate that 66% of the total stu- dents who enrolled in Japanese national universities are male, whereas 71% of students who were enrolled in the uni- versity, where the data were collected, were male. Although the gender distribution of students in the Japanese sample is consistent with the actual gender distributions found in Japa- nese national universities overall, it differs greatly from that of the American sample. Because gender is related to Japan, and also likely related to deviance, gender is a possible source of spurious relationships between nationality and deviance, thus must be controlled throughout the analysis. The respondents’ parental socioeconomic status is mea- sured by combining both parents’ attained education levels. To measure parents’ education levels, students were asked to indicate the highest level of education attained by his or her parent or parents. The variable Parents’ Education is an ordi- nal variable with five categories, coded so that a higher value indicates higher parental education attainment, including ‘‘neither parent has a high school degree’’ (coded 1), ‘‘one or both parents have a high school degree but no higher’’ (coded 2), ‘‘one or both parents have more than a high school degree but less than a college degree’’ (coded 3), ‘‘one or both parents have a college degree but no higher’’ (coded 4), and ‘‘one or both parents have a graduate or pro- fessional degree’’ (coded 5). The distributions of this variable differ by country, and indicate a slightly higher overall paren- tal educational attainment for the American sample com- pared to the Japanese sample. While the percentages of students with at least one parent with a college degree or higher were identical across countries (67.21% for Ameri- cans and 68.29% for Japanese), the percentages of students with at least one parent with a graduate or a professional degree (32.52% and 9.70%, respectively) and the percen- tages of students with both parents having only a high school degree or lower (8.13% and 21.48%, respectively) differ substantially. To measure family structure, students were asked: ‘‘While you were growing up, how would you describe your house- hold?’’ The Two-Parent Home variable in the analysis is a dummy variable. Those students indicating that they had, at any time in their lives, experienced being in a single-par- ent home (i.e., ‘‘single mother household’’ and ‘‘single father household’’) received a value of 0. All other students, that is, those students who had never experienced a single-adult household and lived their entire lives in households with more than one adult present, received a value of 1. The mean for the Two-Parent Home variable for the combined sample is .91 (SD ¼ .29), indicating that 91% of students have never experienced a single-parent household. The distribution dif- fers slightly by nationality. While the mean for the Japanese sample is .95 (SD ¼ .21), the respective mean for the American sample is .85 (SD ¼ .36), indicating a slightly lar- ger proportion of American than Japanese students who have experienced a single-parent home. Social Bonds Measures of social bond elements include attachment to others, commitment to conventional goals, involvement in conventional activities, and belief in the legitimacy of the law. Attachment includes caring about (1) parents, (2) peers, and (3) school. In Hirschi’s original formulation, attachment to, or caring about the feelings of parents has three subcom- ponents: (a) identification with and affection toward parents, (b) intimate communication, and (c) parental supervision. Because of Kempf’s (1993) observation of inconsistencies in how social bond variables have been measured by researchers, we tried as much as possible to use measures identical to those used by Hirschi’s (1969) original research. The analysis includes all four elements of Hirschi’s (1969) social bonds and all three dimensions of attachment. All elements of social bonds are coded so that a higher value indicates a higher level of social bond and missing cases are replaced with the mean. Attachment to parents has three dimensions. Affectional identification was measured with three items (‘‘I want to be like my parents’’; ‘‘I feel extremely close to my parents’’; and ‘‘I get all the affection I want from my parents’’) with an alpha of .81 for the linear composite of z-scores. Intimacy of communication consists of two items, ‘‘I often engage in leisure activities with my parents’’ and ‘‘My parents are will- ing to listen when I need to talk about my worries or pro- blems,’’ whose z-scores are summed to form a scale with an alpha of .72. Finally, alpha was .86 for the sum of z-scores for the two items measuring parental supervision (‘‘How often did your parents figure know where you are when you are away from home?’’ and ‘‘How often did your parents know whom you were with when you were away from home?’’). With the goal of parsimony, we further examined the pos- sibility of creating an overall scale of attachment to parents with these three scales. The result of a reliability analysis for a linear composite of the three, however, could not justify the creation of this single overall scale of attachment to par- ents. Instead, the ‘‘attachment to parents’’ scale is the sum of the z-scores of affectional identification and intimacy of communication (alpha ¼ .87), leaving ‘‘parental supervi- sion’’ as a separate variable (alpha ¼ .86). The attachment to peer scale, with an alpha of .66, is cre- ated by combining the z-score transformation of two items ‘‘Do you respect your close friends’ opinions about the important things in life?’’ and ‘‘Would you like to be the kind of person your close friends are?’’ The single item measuring students’ attachment to school is, ‘‘In general, did you like or dislike high school?’’ High school, rather than college, was chosen as the frame of reference because students in both American and Japanese samples were rather early in their college careers and, thus, might not have solidified views about attachment to college. Commitment is measured by students’ reports of the importance they attach to three goals: getting good grades in college, graduating from college, and attaining a good job. A principal component analysis for these three items indicates a single factor, with eigenvalues of 2.05, .58, and .37. The commitment scale then is created by summing the z-score transformation of these three items (alpha ¼ .75). The involvement variable is measured by the overall time spent in organizations or activities per week. The total of seven types of activities includes working at a job, participa- tion in community organizations, and involvement in school activities such as athletics and band. Students were asked how many hours per week they spent in each of these, and their responses were summed to form the involvement scale. The mean of ‘‘involvement’’ for the combined sample is 22.31 hours per week with the standard deviation of 15.88. The scale for ‘‘belief’’ is created from four items used in Hirschi’s (1969) original study that measure students’ level of agreement=disagreement to statements such as ‘‘it is all right to get around the law if you can get away with it.’’ These items indicate students’ belief in the legitimacy of laws and general rules. The principal components analysis supports a creation of a single factor, with eigenvalues of 2.08, .77, .64, and .52. Cronbach’s alpha for the linear composite of the z-scores of these four items is .69.
ANALYSIS AND RESULTS The analysis begins with the regression of self-reported deviance on Japan to examine whether or not the popular notion of a higher compliance among Japanese compared to Americans would hold in our samples, with controls for age, gender, parents’ education level, and two-parent home shown in Table 1. The direct effect of Japan on self-reported deviance is b ¼ .310, indicating that Japanese score signif- icantly lower on self-reported deviance than do the Ameri- cans. The only other variable with a larger effect is Male (b ¼ .335). Gender often is considered one of the strongest, if not the strongest, predictor of deviance. A comparison of the two coefficients in Table 1 suggests that the effect of being Japanese on reducing deviance is more than 90% as great as the effect of being male on increasing it. In Table 1, the only other variable with a significant direct effect is Age, with a b ¼ .102, suggesting that in the sample TABLE 1 OLS Regression of Deviance on Japan and Control Variables, n ¼ 802 (One-Tailed Significance Tests)
Variable b Beta p
Japan 4.396 0.310 0.000 Male 4.803 0.335 0.000 Age 30.715 0.102 0.001 Parents’ education 0.061 0.008 0.400 Two-parent home 0.532 0.022 0.255 Constant 39.306 R2 0.174 p 0.000 of fairly young college students, Age has a significant positive effect on self-reported deviance. In the equation, 17.4% of the variance of deviance is explained by the five variables, and nearly all of these come from just two variables, Japan and Male. In the next analysis, shown in Table 2, we test our central hypothesis that Japanese have higher levels of social bonds compared to Americans. The results reveal that, contrary to our expectation, the Japanese students do not score consis- tently higher on the social bond variables than the American students. In fact, for many of the social bond variables, con- trary to our expectation from the individualism–collectivism literature, Japanese youth seem to be less strongly bonded to conventional society than are their American counterparts, at
TABLE 2 Partial Correlations between Japan and Social Bonds, n ¼ 802 (One-Tailed Significance Test)
Variables Partial correlations p
Attachment to parents 0.497 0.000 Parental supervision 0.345 0.000 Attachment to peers 0.532 0.000 Attachment to school 0.100 0.002 Involvement 0.257 0.000 Commitment 0.397 0.000 Belief 0.051 0.076 least using our measures. Table 2 reports partial correlations between Japan and each of the seven social bond variables, with controls for gender, age, parents’ education, and two-parent home. Based on the collectivism literature, we predicted these partial correlations to be positive and signifi- cant, suggesting that, with appropriate controls, Japanese are more strongly bonded to conventional society than are Americans. In fact, the only significant positive partial correlation is for Attachment to School. The partial correlation for Belief also is positive but fails to achieve significance at the .05 level (p ¼ .076). For all other social bond variables, Japanese display much weaker bonds than Americans. For Attachment to Parents, Parental Supervision, Attachment to Peers, and Commitment, the partial correlations are not only significant in the direction opposite of expectations, but also they are substantial in magnitude ( .497, .345, .532, and .397). In other words, contrary to our hypoth- esis, Japanese score significantly lower on these social bonds, and the percent of variance in these bonds exp- lained by the Japan dummy variable ranges from 13.5% to 34.1%. The results indicate that the Japanese students engage in a lower level of deviance than Americans, despite their significantly weaker social bonds compared to their American counterparts. We now have only slight evidence that Hirschi’s social control theory could account for the gap in deviance found in samples of the Japanese and the American students. None- theless, the issue of how well each social bond variable explains deviance in the two countries still remains an important question. Thus, in Table 3, we add social bond variables into the original regression equation presented in Table 1. Table 3 shows that with the social bond variables in the equation, the standardized coefficient for Japan in Table 1 of .310 is reduced to .262. In other words, the social bond variables account for only about 15% of the effect of Japan on deviance, net of the control variables, and the effect of Japan with the social bond variables in the equation remains significant. Meanwhile, the social bond variables appear to account for more of the effect of Male (27%) than the effect of Japan, although in Table 2, the effect of Male (b ¼ .244) also remains significant. TABLE 3 OLS Regression of Deviance on Japan, Control Variables, and Social Bonds, n ¼ 802 (One-Tailed Significance Tests)
Variable b Beta p
Japan 3.719 0.262 0.000 Male 3.501 0.244 0.000 Age 21.488 0.072 0.009 Parents’ education 0.004 0.001 0.493 Two-parent home 0.409 0.017 0.289 Attachment to parents 0.056 0.032 0.196 Parental supervision 0.318 0.084 0.006 Attachment to peers 0.117 0.029 0.216 Attachment to school 0.166 0.015 0.309 Commitment 0.060 0.021 0.264 Involvement 0.060 0.134 0.000 Belief 0.925 0.376 0.000 Constant 29.101 R2 0.340 p 0.000
The ability of the social bond variables to account for only 15% of the effect of Japan is disappointing in light of our hypothesis, but expected from our findings in Table 2 that Japanese have less strong social bonds than Americans. The results from the regression analysis with social bond variables further provides evidence that the Japanese stu- dents commit lower level of deviance than Americans despite their lower levels of social bonds. Thus, contrary to Hirschi’s contention, his social control theory fails to account for the differences in levels of deviance found between the Japanese students and American students. It is possible that social bonds failed to reduce the effect of Japan on deviance because all the preexisting variations in the two countries are not accounted for in the analysis. That is, it is possible that the variable Japan not only reflects the cultural variation of collectivism–individualism, thus hypothesized differences in levels of social bonds, but it also subsumes many other preexisting variations between Japan and the United States, such as levels of inequality (Bradshaw and Wallace 1996), availability of firearms (Adler et al. 2004), community policing (Cole and Smith 2001; Thornton and Endo 1992), homogeneity (Kerbo and McKinstry 1995), among other factors. It could be the case that the variable Japan reflects more strongly these other preexisting variations than the difference in collectivism–individualism and social bonds. If so, then even if collectivism–individualism reflects Hirschi’s social bonds, social bond variables by themselves might explain a very small portion of the effect of Japan on deviance. Nonetheless, the fact still remains that although Japanese report weaker social bonds than Americans, they commit lower levels of deviance than Americans. Additionally, the fact that there is not a greater reduction of the effect of Japan on deviance when social bonds are included (in the analysis) can be attributed to two other fac- tors. First, some of the social bond variables (Attachment to Parents, Attachment to Peers, Attachment to School, and Commitment) do not have significant negative effects on deviance. Indeed, one of them (Involvement) actually has a positive effect on deviance. A positive effect of involvement was also found in Hirschi’s (1969) study as well as other stu- dies (e.g., Chaiken 2000; Ploeger 1997). Thus, to some extent, the failure of the social bond variables to account for most of the effect of Japan on deviance can be attributed to the absence of significant inverse effects of the social bond variables on self-reported deviance. Only Parental Supervi- sion (b ¼ –.084) and Belief (b ¼ .376) have significant inverse effects on self-reported deviance, as Hirschi (1969) originally predicted. One possibility we explored, and the second factor that might also explain why the reduction of the effect of Japan on deviance is not greater, is that social bond variables do not have the same effects in Japan as they do in the United States—the issue of generalizability of theory across groups—a possibility that would be masked in Table 3 where the two samples are combined. In Table 4, therefore, we report the effects of each of the social bond variables on self-reported deviance separately for Japanese and Ameri- cans. The coefficients in the table include controls for Male, Age, Parents’ Education, and Two-Parent Home. Using the statistical test discussed in Paternoster, Brame, Mazerolle, and Piquero (1998), the table also includes a test of the sig- nificance of the difference between the Japanese and the American samples in the value of unstandardized coefficient for each social bond variables, which resembles a test of interaction effects. Table 4 shows that in both equations, Attachment to Par- ents and Attachment to School have nonsignificant effects, and the difference between Japanese and Americans is not significant. Parental Supervision has the expected signif- icant inverse effect on deviance among Japanese, while the effect of Parental Supervision is negative but barely fails to achieve significance (p ¼ .055) among the Americans. The difference in the b for Japanese and Americans, however, is not significant. The greatest discrepancy across groups occurs for Attach- ment to Peers. Among the Japanese, Attachment to Peers has a significant positive effect on self-reported deviance (p ¼ .001), an effect inconsistent with the basic premise of Hirschi’s argument but similar to the findings of Hwang and Akers (2003) in South Korea. Among the Americans, however, the effect of Attachment to Peers fails to achieve significance. Furthermore, the difference between the effects of attachment to peers on deviance for the Japanese and the Americans is significant (p ¼ .009). The separate findings for Japanese and Americans for Commitment, Involvement, and Belief mirror the findings for the sample as a whole. Commitment does not have a sig- nificant effect in either group, and the difference in the effect between groups is not significant. For both Japanese and Americans, Involvement has a significant positive effect on self-reported deviance as it did in the combined sample in Table 2. Belief, as was true in the combined sample, has a significant inverse effect for both Japanese and Americans, consistent with Hirschi’s (1969) argument, although the mag- nitude of the effect is significantly less among Japanese than among Americans. In summary, it appears that the failure of the social bond variables to account for much of the effect of Japan on self-reported deviance cannot be attributed to any peculiari- ties in the effects of these variables among Japanese respon- dents. With the exception of the significant positive effect of Attachment to Peers among Japanese, the social bond vari- ables behave remarkably similarly among the Japanese as they do among the Americans, adding evidence to the gen- eralizability of this theory across culture. In effect, with the TABLE 4 OLS Regressions of Deviance on Social Bond Variables in Japanese (n ¼ 433) and American (n ¼ 369) samples with controls for Male, Age, Parents’ Education, and Two-Parent Home (One-Tailed Significance Tests)
Significance of difference Japanese Americans between b’s
Variable b Beta pbBeta pz p
Attachment to parents 0.035 0.018 0.349 0.059 0.029 0.268 0.184 0.429 Parental supervision 0.318 0.099 0.014 0.387 0.074 0.055 0.246 0.401 Attachment to peers 0.483 0.101 0.011 0.233 0.046 0.143 2.376 0.009 Attachment to school 0.018 0.002 0.485 0.490 0.044 0.156 0.758 0.224 Commitment 0.075 0.029 0.258 0.097 0.025 0.277 0.110 0.456 Involvement 0.073 0.176 0.000 0.039 0.079 0.032 1.229 0.109 Belief 0.707 0.291 0.000 1.069 0.452 0.000 2.392 0.008 Constant 43.341 24.017 R2 0.227 0.395 p 0.000 0.000